I want to know if depending on country will an individuals response to 6 different questions which predicts an outcome variable scored as (yes/no) differ. E.g someone from country X may score higher on the 6 questions which in turn predicts whether they answered yes/no to my outcome variable. I'm confused on how to analyze this because if i include country as another IV it wont tell me whether being from a particular country influenced the way respondents answered the 6 questions which in turn affected their outcome response- it simply tell me if country as another IV along with the 6 i already have predicts the outcome.

Could i use the selection variable box on SPSS and run a log.reg. for each country separately? I.e so i would report the logistic model results for each country individually by selecting those cases for country x and again do this for country Y.

  • $\begingroup$ I tried to edit your question to clarify it, but I could not do so as I don't really understand what you a re asking. I think there is a translation problem. $\endgroup$ – Peter Flom - Reinstate Monica Mar 5 '18 at 12:38
  • $\begingroup$ Please check it now, i have clarified it- apologies i wrote it in a rush earlier $\endgroup$ – Ali Mar 5 '18 at 12:48

If you have multiple people from each country and believe that people from one country have some similarity to each other (which seems reasonable) then you will violate the assumption of independent error and regular regression is not appropriate.

What you propose in your second paragraph is called stratification and it is one reasonable thing to do, but you will get separate results for each country with no statistical analysis of the differences.

Another possibility is to use a multilevel model (MLM) . Since your DV is binary, you will need a nonlinear MLM.

  • $\begingroup$ Note- i plan to a run an independent samples t test to find out whether the 2 countries significantly differ on the outcome variables. The main research question here is figuring which of these 6 factors is more important in predicting the outcome yes/no for each country- would i have to statistically analyse the differences here- could i not simply report which variables had an effect for the outcome of each country as this is my main question. I greatly appreciate your time in responding Peter! Thank you again. $\endgroup$ – Ali Mar 5 '18 at 13:04
  • $\begingroup$ Also, as I am only looking at 2 countries would this be sufficient data for a MLM? $\endgroup$ – Ali Mar 5 '18 at 13:13
  • $\begingroup$ I think you want to look at the countries separately. $\endgroup$ – Peter Flom - Reinstate Monica Mar 5 '18 at 13:45
  • $\begingroup$ In a seperate logistical regression model then? $\endgroup$ – Ali Mar 5 '18 at 16:36
  • $\begingroup$ Yes. One for each cointry $\endgroup$ – Peter Flom - Reinstate Monica Mar 6 '18 at 3:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.